We present a study on acoustic modeling of Spanish phonetic units. Bootstrap with a set of English phonetic models, we first obtain context-independent unit models for Spanish. We then compare context-dependent modeling techniques based on the conventional maximum likelihood (ML) and the maximum a posteriori (MAP) criteria. We found the MAP-based context adaptation approach produces a better result than the ML approach when a large number of units need to be modeled but the amount of training data is limited.